Its always fun to try to predict when the next iteration of your favorite toy is going to come out. Since the iPhone 5 is coming in October and Halo Reach is the last iteration to be produced by Bungie, lets talk about OBIEE 188.8.131.52.
In “clue” style, I’m going to go with…. Larry, in the keynote, with a hardware component.
To unpack that a little bit, I think we will either have a GA release of 184.108.40.206 during OpenWorld and/or an announcement of the GA release during OpenWorld. And every year needs to bring a new software-hardware-complete machine, so isn’t it BI’s turn?
Prior keynotes have brought us Exadata (and Exadata 2) and Exalogic. The Oracle database is clearly a pillar of their business, it was a good first choice for a machine. The middleware stack is another large area of the Oracle product line, hence Exalogic. Wouldn’t it be nice if you could run OBIEE on the Exalogic? What would we call an OBIEE machine? ExaIntelligence? The Oracle BI Machine?
Now, just in case I’m right on this, how do I support this? The packaged software and hardware in the “machines” is a way for Oracle to capture revenue that would otherwise go to implementation partners. The BI machine is a logical inevitable progression, if it doesn’t come this year it will be next year. The economics demand it.
On the release of 220.127.116.11, the signs are largely there..
1) google “OBIEE 18.104.22.168 and bug” and you’ll find more and more bugs showing up with notes like “fixed in 22.214.171.124″.
2) the spacing between releases is about right
3) the spacing between releasing is wrong (more on this later)
The 126.96.36.199 release was back in May. Given the increasing indications of progress on 188.8.131.52, you would expect it show up on OTN in August or Sept. This is what happened with the first release of OBIEE 11g. Releasing it before OpenWorld lets the marketing and promotion machine kick into high gear and openly schedule and promote sessions related to the product.
This will be the second OpenWorld for OBI 11g, the first patch set has been released, the BI Apps have been updated to use 11g, early adopters have a year of experience, consultants who had access to the beta releases have 2 years experience. Now is when the majority of clients will begin deciding its safe and make or plan their moves. A pre-OpenWorld release would have generated additional excitement around the product. But there wasn’t one.
The other way to generate excitement is to unveil a surprise. If 184.108.40.206 has support for a machine type solution releasing it early could give that away. This would be a reason to hold it back until the event and announce it with a machine.
What could be in that box?
Qlikview has appeal to some because of the idea of “storing everything in memory”. This sounds really fast, but in practice creates scalability issues. It is challenging to store everything in memory while also scaling out to 10 or 15 machines. Also, a poorly coded algorithm can easily defeat the memory speed advantage. All that aside, Oracle needs to have an in memory answer, leveraging the TimesTen database and a huge amount of RAM or flash would do that.
The BI Application installation, configuration and customization process is fairly involved and time consuming. Providing that service generates a good amount of revenue for service providers (including me). If it could be delivered pre-installed it would allow Oracle to recapture a portion of that revenue. Including this in the first version may be too much of a stretch.
Oracle OLAP and Essbase, which were once competing technologies, are now siblings. OLAP is built into the database but the Exadata machine doesn’t currently have the same benefit on OLAP queries as it does on relational queries. Essbase has the disadvantage of being a separate server and requiring all the data to be moved. An Oracle BI machine would be a good opportunity to at least bring essbase closer to the data. If data must be moved to be used by essbase the high speed InfiniBand connection used in the current Oracle machines would be a large improvement. Ultimately to be effective at scale the essbase server or storage needs to be integrated into the database or the storage cells, similar to the iDB protocol.
Hadoop with its abilities handling “web scale” data has been putting some pressure on Oracle for a while now. With the introduction of the database machine, Oracle managed to gain about an order of magnitude of scale. Hadoop is still well beyond that and has continued to increase in popularity. In large businesses it often finds its place in the ETL phase of BI projects since it lacks the realtime support required for adhoc quires. There will be plenty of Hadoop related talks in the JavaOne part of the OpenWorld conference, it is time to start seeing components of this technology getting included in Oracle’s business products.
The Downside of a Machine
On of the contributions to Oracle’s success in the database market was to ensure the existence of a pool of talent in administering and applying the product. Free developer access via the OTN license was part of this. Any new machines raise the bar for access potentially damage this ecosystem of skill.
In the Exadata machine’s software stack the unique piece is the iDB protocol used to communicate between the storage cells and the database server. The iDB component is not available except with the purchase of an Exadata or storage cell. It would help expand the talent pool to have an OTN article series showing how to build a RAC using storage cells on 3rd party hardware and Oracle Enterprise Linux. Hopefully, there won’t be a similar component that is required but not generally available when there is an Oracle BI Machine.
OK, that my case. Whats yours? When do you think 220.127.116.11 will be released? Will there be a new machine announced and if so what will it focus on?